"AI systems are just text samplers": dev argues intelligence lives in users, not models
gerardsans · x · 2026-11-18
Developer Gerard Sans argued on X that if AI outcomes depend entirely on how well users prompt, the difference lies in the user, not the system — so the system provides no intelligence and instead consumes it. His interlocutor pushed back that this is like denying highly intelligent people exist. Sans responded that his opponent is projecting human traits onto AI: current systems are text samplers with no intelligence or human traits whatsoever, which may be the source of the misunderstanding.
More from AGI Musings
- 'AI can't transfer expertise because it has none': a spicy AI skeptic take — gerardsans · 2026-11-18
- "Current Systems Are Just Text Samplers": gerardsans Doubles Down in AI Intelligence Debate — ziv_ravid · 2026-11-18
- Enter "Cyborg Code": Humans Writing With AI Means Fewer Bugs, More Resilience — technollama · 2026-11-18
- CDC data shows US youth sadness and suicidality falling sharply since 2021, undercutting phone-crisis claim — skdh · 2026-11-18
- Friend creator Avi Schiffmann rants against screens, calls for trials of Meta and Apple — ctjlewis · 2026-11-18
- KPMG Deployed 100+ AI Agents in a Real Audit and Found They Asked Better Questions Than Humans — The Decoder · 2026-11-18